AI agents with OpenAI webhooks

Receive OpenAI webhooks on Trigger.dev and use them to start, resume or message your AI agents.

Get started now

What you can do with OpenAI webhooks

Start an agent

OpenAI events can start a new agent run. For example, an agent that picks up results when an OpenAI batch completes.

Resume an agent

An agent can pause until the OpenAI event it needs arrives, for up to 90 days, and costs nothing while it waits.

Message an agent

OpenAI events can land in a running agent's conversation, so it acts on them with full context.

Events you can receive from OpenAI

Some of the events OpenAI sends. Your agent can filter on any of them.

batch.completed

Batch completed

batch.failed

Batch failed

fine_tuning.job.succeeded

Fine-tuning job succeeded

response.completed

Response completed

eval.run.succeeded

Eval run succeeded

realtime.call.incoming

Incoming Realtime SIP call

Example

An agent that picks up results when an OpenAI batch completes

import { webhook, webhooks } from "@trigger.dev/sdk";
export const openai = webhooks.endpoint.define({
id: "openai",
source: webhooks.openai(),
});
export const openaiEvents = webhook({
id: "openai-events",
endpoint: openai,
filter: "event.type == 'batch.completed'",
onEvent: async ({ event }) => {
// Start, resume or message your agent with the event
console.log(event);
},
});

Handled for you

Verified for you

webhooks.openai() checks the OpenAI signature on every request. Bad requests are rejected before anything runs.

Retries don't run twice

Retries are matched on the webhook-id header, so a retried webhook doesn't start a second run.

Filter before anything runs

Only the OpenAI events you care about start a run. Filtered webhooks use no compute.

See every delivery

Every webhook is recorded with its payload, headers and the agents it reached, so you can debug it and replay it.

Get started

  1. 1

    Set up the agent

    Copy the starter prompt into your coding agent, like Claude Code, Cursor or Codex. It creates a Trigger.dev project and adds the webhook endpoint. Prefer to write it yourself? Start from the example above.

  2. 2

    Add the webhook in OpenAI

    Deploy your project, or run trigger dev, and the endpoint appears on the Webhooks page of the Trigger.dev dashboard with its URL. Add that URL in OpenAI's webhook settings and pick the events you need. The endpoint's Copy setup prompt button gives your coding agent the exact steps.

  3. 3

    Add OpenAI's signing secret

    Copy the signing secret from OpenAI. Then on the endpoint's page in the Trigger.dev dashboard, choose Set signing secret and paste it in, so Trigger.dev can check every request really came from OpenAI.

  4. 4

    Try it

    Send a test event from OpenAI, or use the endpoint's Test tab. It appears on the endpoint's Deliveries tab, with the agents it reached.

Related integrations

Ready to build with OpenAI webhooks?

Declare an endpoint, deploy, and paste the URL into your provider.

Get started for free